(4.1.7), p. 43 — the squared norm of a matrix Gaussian series
ProvedIntroMatrixConc.GaussianSeries.eq_4_1_7matrix-concentrationp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1random-matrices
Let be a finite family of fixed complex matrices, where , and let the be independent standard real Gaussian variables. Set and . Then is integrable and
The result identifies the matrix variance as the scale of the squared operator norm, allowing the logarithmic dimension factor on the upper side.
Formalization Note Integrability is stated explicitly so that the Bochner integral cannot take Lean's default zero value for a nonintegrable function. An empty coefficient family is allowed and gives .
Preamble
import Mathlib import Definitions.Def_TroppMatrixConcentration_ch4_scalar_laws open MeasureTheory ProbabilityTheory TroppMatrixConcentration open scoped Matrix.Norms.L2Operator noncomputable section
Formal statement
namespace IntroMatrixConc.GaussianSeries
/-- Tropp (4.1.7): both sides of the squared-norm comparison for a Gaussian series. -/
theorem eq_4_1_7 {Ω : Type*} [MeasurableSpace Ω]
(μ : Measure Ω) [IsProbabilityMeasure μ]
{d₁ d₂ N : ℕ} [NeZero d₁] [NeZero d₂]
(B : Fin N → Matrix (Fin d₁) (Fin d₂) ℂ)
(g : Fin N → Ω → ℝ)
(hMeas : ∀ k, Measurable (g k))
(hIndep : iIndepFun g μ)
(hLaw : ∀ k, standardGaussianLaw μ (g k)) :
let Z := fun ω => ∑ k, g k ω • B k
let v := rectSecondMoment μ Z
Integrable (fun ω => spectralNorm (Z ω) ^ 2) μ ∧
v ≤ ∫ ω, spectralNorm (Z ω) ^ 2 ∂μ ∧
(∫ ω, spectralNorm (Z ω) ^ 2 ∂μ) ≤
2 * v * (1 + Real.log (d₁ + d₂)) := by sorry
end IntroMatrixConc.GaussianSeries
Source
Tropp, An Introduction to Matrix Concentration Inequalities, arXiv:1501.01571v1, (4.1.7), pp. 43–44
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.